模型与 AI 能力

Use your ChatGPT subscription in DeepSeek Harness. Sign in, choose GPT, and work with imag

Provider-neutral change request registry with shared status polling

A second brain for your DeepSeek Harness agent — strong-model review before risky tool cal

OpenAI Codex OAuth provider and model adapter for dsh-oauth and DeepSeek Harness

Continuous voice conversations in your current DeepSeek Harness chat · self-hosted Breeze

Two-tier model router for DeepSeek Harness: an LLM Judge sends each turn to the cheap Fast

把天津大学智算平台（ai.tju.edu.cn）接入 DeepSeek Harness（dsh）——社区工具，非官方

DeepSeek Harness 生图插件：文生图 / 图片生成，基于火山方舟 Seedream（豆包）。零依赖，纯 JavaScript，无需构建。 | DSH image ge

Declare reasoningEfforts and vision modalities for hand-written dsh provider routes the pi

Configure model input capabilities and reasoning efforts for DeepSeek Harness

自动保持 profile 内 @deepseek-ai 插件版本与 DeepSeek Harness 运行时一致：应用升级后自动对齐配套插件，安装失败或装到旧版时按退避策略自动重试

Image and video generation for DeepSeek Harness: routes image models to /images/generation

Owner-only task model configuration API for DeepSeek Harness, starting with session titles

鲸尾网关助手：DeepSeek Harness 移动客户端连接与会话同步，支持安全配对、远程访问、上下文、编队、额度和语音输入

Image generation through the locally installed Doubao desktop app, driven over the Chrome

DSH plugin: read images through a user-configured vision-capable model when the main model

Local real-time speech-to-text for DeepSeek Harness: a floating transcription panel plus a

Service Provider for image generation in the DeepSeek Harness, over the OpenRouter Image A

Model and reasoning effort picker for the DeepSeek Harness composer.

Natural, human-sounding voice for the dsh harness: server-proxied neural TTS read-aloud an

Per-model capability declarations (image input and reasoning efforts) for custom pi-ai pro

《NSCA-CSCS体能教练认证指南（第4版）》学习插件（DeepSeek Harness/dsh 版）：注册 nsca_ocr_page / nsca_use_book 工具。不

Give DeepSeek Harness teammate (continuable subagent) sessions their own model picker: swi

LLMTR AI gateway provider for the DeepSeek Harness LLM seam: 170+ chat models and Türkiye-

Hold Space to talk, release to auto-send - push-to-talk for DeepSeek Harness voice input.

DSH model provider for Cohere — OpenAI-compatible wire by default, Cohere's native v2 lane

TG给D.H发语音。光速识别 超迅捷识别。秒回。你懂的。一般情况下只支持打字。Telegram voice -> local sherpa-onnx transcript for

Session-scoped Standard/Fast mode and a compact draggable model picker for DeepSeek Harnes

Screenshot button for the DeepSeek Harness Web UI 鈥?region capture, annotation (pen/line/b

DSH plugin: auto-route to a multimodal model when messages contain images, stay on the che

DSH plugin: expose Trae Work CN credits as a local OpenAI/Anthropic compatible API for Cla

Voice plugin for DeepSeek Harness (dual-face): cloud ASR (Doubao SAUC) + speechified TTS p

DeepSeek Harness native plugin: vision capabilities for text-only LLMs (describe, OCR, VQA

Free OpenCode Zen models for DeepSeek Harness (DSH): native dsh-llm adapter plugin, no API

Qwen, DeepSeek, Kimi and Z.ai web chats as DeepSeek Harness model providers, signed in thr

Portable DSH bundle, profile, and Hub API schemas

Ocr plugin twin - completing the pair - name reserved; first release in development.

Model Ultra — DSH plugin for llm-pi-ai provider lifecycle: create/edit, custom headers, en

DeepSeek Harness plugin: live on/off switch for the Harness outbound HTTP(S) proxy — setti

DSH plugin: a virtual model-router provider that aggregates every configured provider and

Fully-local OCR CLI for text-only LLMs: PaddleOCR-VL first-tier engine with tesseract fall

Screenshot editor that lands the crop straight in the composer draft.

DSH plugin: fill custom-provider model metadata from models.dev through an explicit, pre-d

DeepSeek Harness (dsh) plugin: turn one idea into a narrated video cut from your own asset

DSH plugin: image understanding via any OpenAI-compatible vision API, paste-to-describe, a

First-class LocalAI provider for DSH language, embedding, reranking, decision, speech, and

Installable Orb bundle. Inserts the Computer Use preset and the floating-ball host.

DeepSeek Harness 插件：在 DSH 里管理 CLIProxyAPI 的账号 —— 余额、签到、任务、切换账号，并让 CPA 随 DSH 启停。

通过 WorkBuddy API Key 或 WorkBuddy 登录令牌为 DeepSeek Harness 接入 WorkBuddy 模型

WorkBuddy (via the local workbuddy2api proxy) LLM provider adapter for DeepSeek Harness, w

A persistent, human-approved long-form novel writing plugin for DeepSeek Harness

Auto-continue plugin for DSH: continues model responses truncated at max-tokens

Reusable System One decision transport for DSH AI providers.

DSH plugin for Agnes AI text-to-image generation

dsh plugin: verbatim chapter archive + zero-token index for small-context models

DSH plugin: split-panel model and reasoning-effort picker for the composer.

Serve portable DeepSeek Harness (DSH) plugin capabilities as an MCP stdio server for CLAT

DSH plugin: expose TypeSafe Jev (System One decision model) as the model-callable jev_deci

DSH plugin: configure per-model reasoning effort levels for custom pi-ai providers in the

Local speech transcription DSH plugin (open source): TranscriberBackend seam with MLX impl

DSH plugin: connect to MCP servers, discover tools, and register them for direct model inv

Direct-API image generation DSH plugin (open source): GPT-image behind one image_generate

DSH plugin connecting the agent to MCP servers through one mcp proxy tool and a Settings >

DSH plugin: refreshes the model catalog, probes every model for availability and separates

DSH plugin: wait_for_condition tool — pause and wait for an external action to complete, o

Langfuse observability for DeepSeek Harness: OpenTelemetry traces, compaction, feedback Sc

OpenAI language, embedding, decision, speech, and image provider for DSH.

Drive DeepSeek Harness (DSH) as a long-lived coding agent from any MCP client — Claude Cod

BotHouse channels for DeepSeek Harness: Discord-style rooms where DSH models and local CLI

DSH 模型补齐助手：解决添加新渠道后还要把几十个模型一个个手动点开调百万上下文、调最大输出、勾多模态发图（\[text, image]）、配思考强度档位（{low, high,

Ollama Cloud 供应商：安装即配好聊天路由，按 Ollama 模型自身元数据公布每个模型的思维链档位（供 composer 的推理强度选择），模型清单可由端点自动刷新（/

为未声明思考档位的模型编辑推理强度档位与思考格式。

为 Windows 上的 DSH Desktop 0.2.0-rc.2 提供原生 Codex OAuth 登录、主 Agent 默认模型选择及直连或本地代理设置。

按 Provider + Model 记住你最后一次明确选择的推理档位，切回该模型时用同一次原生选择调用恢复，不替换 DSH 原生模型选择器。

把 Factory (Droid) 订阅额度接入 DSH 的原生模型 provider：三条协议路由、用你自己的 API key 经本机回环网关认证，并提供设置页查看连接状态、订阅

将你自己的 Hugging Face GPU Space 中的模型加入模型选择器，发送消息时启动该 Space，显示启动进度，并在空闲或达到时限时将其暂停。需要 Hugging F

需要 DSH >= 0.2.0-rc.2 < 0.3.0。在设置页配置任意数量的模型供应商：各自独立设置接口地址、密钥与逐模型参数（思考强度、图片输入、重试策略），支持 OpenA

把 WorkBuddy 国内版和国际版的免费/付费模型接进 DSH —— 无需下载 WorkBuddy 桌面端。

将 TRAE 账号的模型额度接入 dsh，支持多账号管理、自动签到、账号轮询、用量统计、思考强度选择和模型显示管理。 需要 Python 3.10+，首次使用需联网安装 Pytho

复用 Responses 存储 ID 续传，隔离重试与分支，并核对历史回放。

通过内置 pi-ai 适配器将 Standard Compute 添加为 OpenAI 兼容模型供应商，包含 standardcompute 路由模型和两个直连模型；仅配置，无代码

把 chat.deepseek.com 网页版接成 deepseek-web provider：纯 HTTP 无浏览器，多账号轮换/冷却/自动切换，插件内登录（浏览器，或邮箱+密码

为 opencode-go 与 Command Code 路由声明 wire 协议和 API 地址（不用手填地址）；把 DeepSeek V4.1 模型补进 opencode-go

模型选择器用的可配置降级链：有序主链加命名池，按失败码自动切换，支持按模型冷却、分时段预设与子代理角色路由。omdsh-dev/dsh-llm-fallbacks 0.6.5 的

在设置里新增「模型能力」一页，为每个模型声明思考档位与视觉能力。

DSH 设置页里的百炼（DashScope / Model Studio）模型目录：列出当前 API Key 可用的全部模型、按模型家族分组，可探测某模型是否真的可调用（未开通会返

大模型请求重试插件：命中配置的 HTTP 状态码、机器码或 provider 返回字段=值时，按该条规则独立配置的时长休眠后自动重新请求，让订阅套餐的限额、限流报错（429 / i

模型选择弹窗＋路由分组：模型失败后自动换到组内下一个可用模型并继续本轮；另含数据看板与免费渠道页签。页签一键安装六个渠道插件 —— Our Free Model（唯一不登录、不填

通过 ACP 将腾讯 CodeBuddy CLI 接入 DeepSeek Harness，支持 CLI 原生工具执行、多账户管理和 codebuddy_ask 独立问答。

通过 OAuth 将 ChatGPT 订阅模型接入 DSH，并在会话回放中保留本地工具调用身份。

通过官方 Qoder CLI 以原生 ACP stdio 接入 Qoder 模型（无需逆向模型 API）；qoderclicn 中国版已通过真实 ACP 工具链路验证，qoderc

通过 OAuth 接入 Claude Pro/Max 订阅，支持多账号、可选故障切换、额度缓存、在线模型目录，以及设置页和会话额度入口。请求使用 Claude Code 客户端身份

本机已登录的 Codex（ChatGPT Plus/Pro）订阅接入 DSH。用途：①在侧边栏展示 5 小时窗口与一周窗口的剩余额度、重置时间与剩余重置券数量，并支持一键消耗一张重

按模型声明的能力提供思考深度滑块、配置指引与可选模型行设置。

DeepSeek Harness 的模型故障自动转移：可选一个或多个供应商，池=所选供应商内的全部模型；请求失败（限流、配额、超时、服务端或传输错误）时自动用池中下一个未试过的模型

可选 JEV 决策模式，提供代码优先项目读取、有限语义选择、原生授权检查与桌面密钥设置。

多账号 Google Gemini 提供商：按剩余额度挑选账号，遇 429 指数退避切换，后台探活已禁用账号，附中英双语设置页。

在 DeepSeek Harness 的模型目录里识别模型 ID，用内置的离线规则表（覆盖 33 家厂商）把该行的上下文窗口、输入模态与思考等级填成可修改的建议。

接入 SAP AI Core，共用 OAuth 认证，支持五个 Qwen 模型、五档思考预算及关闭选项、流式输出和工具调用；已实测 cn40。

原生多平台账号与模型路由管理，支持错峰签到、事件通知与额度预测。后台 worker 需独立 Node ≥22.23.3（node:sqlite，已由 engines.node 约束